Every tech company you can think of is jumping on the generative AI bandwagon and touting new features promising to make our lives easier, increase productivity, and unlock some dormant cache of hidden potential within all of us. 

But “promise” is the operative word here. Despite all the AI hype and billions of dollars of investment, generative AI is still very new to the average person and has yet to transform from being a fascinating novelty into an indispensable mainstay. 

Even as someone who writes about AI for a living, I use generative AI tools somewhat sporadically. That mostly involves testing out and experimenting with new AI models and apps that are available. I might get excited about a certain app or tool only to forget about it and fall back into my normal habits.

Also, I generally don’t trust generative AI. Since truth and accuracy is pretty important to what I do, bringing ChatGPT and other tools into the fold seems not only counterproductive to me, but also potentially disastrous for my employment. 

But what if there are ways of using AI tools that don’t replace my livelihood, but empower me to do it better? Isn’t that what Big Tech is promising us anyway? That’s why I decided to give it the old college try and spend some time using AI tools in earnest to test out what’s useful, and more importantly what actually sticks. 

The experiment 

I spent a little over a week using generative AI tools that fit within my daily life and work schedule. To do this, I made an outline of what my typical week looks like and identified ways where generative AI could help and which tools to use. 

Since several AI tools require a subscription, programming knowledge, or aren’t widely available yet, I narrowed it down to services that are free or readily available to the average person. That excludes tools like Slack’s AI add-on, Microsoft’s full access to Copilot, and Zoom’s AI companion which requires a paid subscription.

The experiment isn’t intended to use as many generative AI tools as possible, but to experiment with the popular ones deemed the most useful for my life. 

Here’s what I came up with: 

It’s important to note that I have a paid subscription to ChatGPT (ChatGPT Plus), and Mashable paid for a subscription to Otter. But I used all of the other apps with my free personal accounts.

First, a disclaimer

Being a reporter, using generative AI tools for work comes with unique ethical challenges. That’s why I pledged to use these tools only for brainstorming and writing help, but never to actually write something for me.

Anytime I used generative AI to help with drafting or brainstorming, I did it for the purposes of understanding whether it was actually useful and accurate. So you’ll see examples of me asking for writing-related help but I didn’t actually publish or post any of its responses verbatim. 

Also, when using ChatGPT for any unpublished or sensitive information I had my chat history turned off. This was to ensure that what I shared wasn’t used to train the model. I didn’t use Gemini or Copilot in these instances because there’s no way to protect your data from being used, unless you have a paid enterprise account. 

With expectations firmly set, let’s begin!

Spinning my tangled to-do list into efficiency gold

Organization and time management are not my strong suit. I seem to be missing that gene that accurately budgets time, plans ahead, and efficiently doles out my time and energy instead of obsessing over mundane details and doing everything at the last minute. 

Here’s where I was most excited to lean on AI tools. Every morning, I started my day by brain dumping all the things I needed and wanted to accomplish that day. This included work tasks related to story ideas and assignments and life tasks like walking my dog, doing laundry, and doing yoga. 

On the first day, I fed my to-do list with specifications around my work hours, any imminent deadlines, and appointments into ChatGPT, Gemini, and Copilot to see which one gave me the most helpful response. It became immediately clear that ChatGPT was to become my go-to for the rest of the week. 

Even though I told Gemini I worked from 9 a.m. to 6 p.m., it had me doing work tasks before 9 a.m. and working up until 9 p.m., without any specifics on when to accomplish certain tasks, only suggesting I should “swap the yoga and writing session in the evening if that works better for you.” That might be how the workday goes in Silicon Valley, but I keep pretty strict work hours to stay sane, especially when working remotely. 

I think Gemini needs to learn more about work/life balance.
Credit: Google

Copilot was better at actually blocking out my time, but since it kept encouraging internet browsing by offering me links to various articles about where to find yoga studios or links to articles about AI tools (since I explained what I was working on.) 

Copilot’s response wasn’t bad but I didn’t like being prompted with links in this context.
Credit: Microsoft

ChatGPT, on the other hand, gave me a genuinely useful schedule based on how to best use my creative energy. After tackling a news story to start the day, it suggested a concrete time block for my longer term story in the early afternoon, since “It’s a good period to focus deeply on research, outlining, or writing without the fatigue that can come later in the day.” It even recommended getting the laundry going before I work in my longterm story since it can run in the background. I was going to do this anyway, but it was interesting to see ChatGPT picking up on that efficiency. 

ChatGPT gave me the no-nonsense, efficient schedule I needed.
Credit: OpenAI

Toward the end of the day, ChatGPT suggested less energy-consuming tasks like reaching out to new sources. 

With my ChatGPT-planned schedule in place every morning, I felt more organized and relaxed about the day. Some things definitely took longer than ChatGPT budgeted time for, but it was useful to have something to stick to instead of my usual vague to-do list. 

Good with recaps, bad with headlines

Throughout the workday, I looked for ways to incorporate generative AI into my work. Since I definitely wasn’t asking it to write for me, I had to get creative about how else to use it. That’s how I ended up using ChatGPT for its summarization skills.

When struggling with a headline, I pasted a draft of the story into ChatGPT and asked it what the story was about. This helped me make sure what I was trying to communicate lined up with what I wrote, and determine certain keywords to use in a headline. 

Just as an experiment, I also asked ChatGPT to give me headline ideas using stories I’d previously written to see what it came up with. It accurately captured the gist of the story, but ChatGPT’s headlines were predictably sterile and robotic. 

For my story regarding, OpenAI’s lack of transparency around training data for its AI video generator Sora, ChatGPT suggested headlines like “OpenAI’s Sora: Breakthrough Video AI Sparks Copyright Controversy” and “Sora Unveiled: OpenAI’s Latest Leap Forward Amid Ethical Quandaries.” Putting aside the fact that ChatGPT loves a dramatic colon, it sounds like a poor imitation of a New York Times article.

None of these headlines were particularly inspiring.
Credit: OpenAI

While Gemini’s headline ideas were either very generic like “The Legal and Ethical Issues of Training AI Models on Public Data” or in a clickbait-y style that makes me dread the future of the internet, like “Is OpenAI Stealing Data to Train its AI Models?” 

I also pasted the same story into Copilot for headline ideas, but it was cut off because there’s a 2,000-word limit. 

Encountering the trust issue

Now onto the next chapter of my day. This week involved reading through transcripts of interviews for a longer term story I’m working on. For journalists, Otter is a lifesaver. It uses AI to transcribe audio files so you can review your interviews within minutes instead of tediously transcribing them yourself. 

That’s nothing new, but this year, Otter released an AI chatbot, which also provides transcript summaries, tells you if you were mentioned, and outlines any specific takeaways. This is mostly geared towards workers who use Otter for meetings, not for journalists interviewing sources, but I decided to try it out and see if it helped for my purposes, since transcripts can be burdensome to comb through for salient points. In the Otter Chat for a long interview, I asked “what are the takeaways from this conversation?” The results helpfully showed a brief bulleted list of the major points of the conversation. 

The AI-generated takeaways seemed valid, but I still felt like I needed to read through the transcript and double-check, since we all know at this point, generative AI can sometimes get it wrong. Everything ultimately was correct, but this is an example of where generative AI might not be the time-saver it’s cracked up to be if everything needs to be fact-checked. Nonetheless, I liked having the list of takeaways as a reference when going back to the transcript instead of having to scroll through all of it. 

Wrangling customer service is an AI solution I can get behind

I wanted to try out Google’s AI tools for Gmail (formerly Duet AI, now rolled into the umbrella Gemini brand) but only had access to it with my personal email through Google Labs. During my lunch break one day, I discovered something beautiful that I will definitely be using in the future: emailing customer service. 

As a Soda Stream devotee, I’m a big fan of their CO2 exchange, which lets me ship back used canisters in the same box that delivered the new ones, all with a prepaid label. Yet, when I ordered new canisters, the delivery date came and went, with no new canisters to make my water delightfully carbonated. The tracking status said they were delivered, but nothing ever arrived. 

These are the kinds of mundane tasks I despise. Every time I remembered to reach out to Soda Stream, something else would come up and it would slip my mind. I also debated whether it would be better to email them or call, which would mean sitting on hold, and then would I have to somehow prove my canisters never arrived? Welcome to my neurotic brain. 

Bolstered by the spirit of my experiment, I turned to Gmail’s “Help me write.” In the chat, I prompted “draft an email to Soda Stream explaining how my CO2 canisters never arrived, even though it said they were delivered on January 9.” The result was a perfectly adequate email, and subject line that explained all the necessary details. All I did was add in the order number, since I didn’t specify that in the original prompt. 

Now this is helpful.
Credit: Google

By the way, Soda Stream was very responsive and sent me new canisters right away, no questions asked. All in all, it was a pain-free experience, and now I’m back to drinking carbonated water by the gallon. 

Half-baked recipe ideas

After my workday, I put generative AI tools to the test while cooking dinner. Tasty’s Botatouille app is a chatbot from the popular recipe app that helps you brainstorm meals and find recipes. 

At the end of most days when my creativity is tapped, I struggle with making sense of the contents of my fridge, let alone how to cook something. So I started out by asking Botatouille for ideas based on the ingredients I had. In terms of generating meals, Botatouille had some good ideas. One night, I made a soup with kale, white beans, and sausage, and another night, I threw together a stir fry with leftover veggies and chicken. 

However, Botatouille was better as a jumping off point for finding ideas rather than executing the recipe. I often had to follow up to get the recipe, and even then I ended up reworking it based on my own needs and ideas. I tend to improvise or add my own flair to recipes when cooking anyways (which is why I’m a terrible baker). 

But that was the extent of its usefulness. One major reason is because Botatouille doesn’t save your chat history, so I couldn’t go back and revisit ideas. I was also disappointed when looking for recipes using ground beef and shiitake mushrooms, since it kept giving me ideas for other types of meat, which I did not have in my refrigerator. Maybe if you’re planning on cooking a meal for a dinner party and going out to buy ingredients, it might be more helpful, but for a hungry person who just wants to eat and relax, the best way to use it is as a brainstorming shortcut. 

Botatouille gave me one recipe idea with ground beef, but the other two used different types of meat.
Credit: Tasty

In conclusion…

My week with AI tools was a bit of a mixed bag. Having committed to using them as much as possible throughout my day, there are a few tools I could see myself coming back to and some that don’t have the stickiness to earn a spot on my tech roster. 

I’ll likely keep using ChatGPT for planning my workday and summarizing information, and brainstorming headlines. I will also be relying more on Gmail’s “Help me write” feature for tackling any sort of mindless email-related tasks that just need to get done. 

But some of the other tools felt a bit redundant or required more effort than they were worth. Otter’s AI companion tools didn’t save me much time since I felt the need to double-check its responses. With Tasty’s Botatouille, I could have just googled recipe ideas or browsed through the app, rather than go through the extra step of writing out a prompt. 

Ultimately, many of the AI tools for work are geared towards workers in non-creative industries. I could see a project manager or consultant, whose days are jam-packed with meetings and involve lots of logistics and task-juggling, relying on generative AI to catch them up on meetings, put together recaps of strategic conversations, and design proposals and presentations.

But, since most of what I do involves teasing words out of the old brainbox, there’s only so much ChatGPT and the like can help with, and I’m pretty okay with that.